A Research On The Methods For The Extraction Of Human Iris Feature

Iris recognition is a highly efficient way of biometrics. An automatically capturing iris system on the basis of image quality appreciation is introduced. The iris image of high quality is then filtered by median value. A value threshold is found from the iris histogram to separate the pupil area.And after the continuous area detection, the eyelashes noise is eradicated.The pupil is located in a way searching coarsely the pixel,in whose line and column the total pixels value sum is maximal.And then integral-differential operator precisely determine the inner center.In order to decrease noise disturbance,according to the inner parameter,Daugman's area integral calculus is adopted to detect the iris outer boundary.And then homomorphism filtering enhances the texture details.After the normalization of the iris area,2-D optimized Gabor transform and zero crossing are performed,after which the image is dealt with mathematic morphology to remove noise.Wavelets analysis can resolved signals and images in different scales.So it is widely used in pattern recognition and feature extraction. In this paper, the iris feature extraction and encoding is done by 2D wavelet transform and coefficients zero-crossing. In contrast to Daugman's algorithm, both memory and time are saved.The match rate is defined by Hamming similarity according to the matched codes for two images to judge whether two images is from the same eye. After experiments on 44435 pairs of iris, the classification has got the accuracy rate of 97%. We analyze the existing and proposed algorithms to advance the next work and improve the algorithm.